Cybersmish: A Proactive Approach for Smishing Detection and Prevention using Machine Learning
W. L. T. T. N. Kumarasiri, M. K. J. C. Siriwardhana, S. A. D. S. L. Suraweera, Amila Nuwan Senarathne, S.M.B. Harshanath · 2023
Social engineering attacks pose a significant risk to user data confidentiality, including credit card details and social network passwords. To combat the rise of smishing attacks targeting digital devices, this research proposes Cybersmish, which utilizes machine learning and NLP Transformer-based models. Cybersmish’s three main components include a message content analyzer, URL analyzer and HTML content analyzer. Leveraging pre-trained models such as MobileBERT, DistilBERT, XGBoost and TF-BERT, Cybersmish achieves accuracy levels exceeding 90% in detecting and analyzing SMSs, classifying URLs, and evaluating HTML web application authenticity. This research underscores the effectiveness of deploying advanced technologies to enhance mobile device security, protecting users from social engineering attacks. It addresses the need for robust solutions against evolving cyber threats, emphasizing the critical role of technology in safeguarding sensitive information in the digital age.